Search

There is increasing evidence that the composition of the human gut microbiome has a key role in health. Environmental factors predominate over the genetic in shaping the microbiome, yet while the potential factors contributing to the extensive inter-individual differences of microbiota are numerous, many have yet to be analysed. This thesis provides exploratory evidence of the relationship between wider environmental variables and microbiota composition and variance using the national TwinsUK cohort. The first part of this thesis provides a conceptual framework for the environmental factors potentially impacting the gut microbiota comprising four broad domains: hostintrinsic, host-lifestyle, passive and distal. Then, selecting key tractable variables from each conceptual domain, observational methods are used to analyse their association with microbiota composition characterised using 16s rRNA gene sequencing data from the TwinsUK cohort. Previous studies have focused on host-intrinsic (e.g. genetic, host health status) and host-lifestyle (e.g. diet, drugs) factors. Diet has emerged as a key environmental determinant, but the best method to capture the overall contribution of diet to variance in microbiota was not known. Therefore, in the second part of this thesis three dietary indices are compared, demonstrating that the 'Healthy Eating Index' captures the majority of variance attributable to diet. Next, the focus of the thesis is to provide examples of underexplored environmental factors, hypothesised to be associated with microbiota composition; distalenvironmental factors and passive-environmental factors. The models used to analyse each of these factors incorporated key covariates related to host intrinsic and host lifestyle factors. The passive environment captures inadvertently consumed factors that are products of the distal via host-lifestyle, such as food contaminants. The passive environment is explored via the composition of tap water, and results suggest, but are not conclusive of, an association with microbiota composition. Future study of passive environmental factors is therefore merited. Finally, the distal environment is examined. Distal-environmental factors are those associated with the wider geosocial and spatial environment of the host. Here, it is explored via socioeconomic status and the host's habitat; both studies demonstrate these understudied factors to have significant associations with multiple measures of the microbiota. These findings raise the possibility that the microbiota mediate the effect of distal factors on host health. In summary, within this thesis a conceptual framework is used to explore understudied environmental associates of microbiota variance. Although not comprehensive, the examples studied provide evidence that distal and passive environmental factors merit future exploration. The results are of interest not only in furthering understanding of influencers of microbiota composition, but also in contributing to knowledge of how an individual's spatial environment interacts with their health.

There is an increasing body of research associating the gut microbiome with human health. This has been fuelled by the development of untargeted techniques to profile whole microbial communities. A common approach is to sequence the 16S rRNA gene as a marker for taxonomic quantification. However, the results of such analyses are of limited use. The high-dimensional sequencing data are typically reduced to operational taxonomic units (OTUs), which are purely analytical units with no direct biological interpretation. Furthermore, experimental and analytical variation can lead to poor reproducibility of results between such studies. Here, I present novel methods to address these limitations and improve the application of 16S rRNA gene sequencing to profiling of the gut microbiota in human health research. I first explore existing and simpler approaches to the analysis of 16S rRNA gene sequencing data by investigating gut microbiota associations with two phenotypes, host frailty and proton pump inhibitor use, identifying several novel associations with both. I then tackle the limited biological representation of OTUs by presenting a comprehensive comparison of OTU clustering algorithms, using heritability as a novel and biologically motivated quality measure. The results of this comparison provide guidance for the approaches used in later analyses. Finally, I present two novel, more complex, methods to summarise 16S rRNA gene sequencing data in human health studies; both of which address the high dimensionality and limited reproducibility and biological relevance of these datasets. Firstly, I identify gut microbiota associations with multiple diseases within a single study. This generates comparable results that enable the identification of key marker taxa consistently associated with health or disease. From this, I generate an index that can represent widescale gut microbiota composition within a single score that is also associated with host health. Secondly, I present a method to identify communities of interacting microbes within the gut microbiota and demonstrate that these reliably associate with host phenotypes across geographically diverse populations. In summary, this thesis addresses several issues currently associated with the analysis of 16S rRNA gene sequencing-based gut microbiota profiles and in turn identifies several novel biological phenomena. The techniques described herein should improve the utility of microbiota profiles derived from 16S rRNA gene sequencing and advance the field of gut microbiome research in human health.

Epigenetic mechanisms respond to both genetic and environmental factors, but these underlying effects are not yet fully characterized or completely understood. In this thesis I used twins as a tool for evaluating the effect of genetic and environmental impacts on DNA methylation, a well-known epigenetic mechanism. DNA methylation was studied on a genome-wide scale in human blood samples with a combination of bioinformatics, computational, and statistical approaches. First, genetic influences on DNA methylation profiles were assessed by estimation of the heritability of DNA methylation and identification of methylation-quantitative trait loci in identical and non-identical twins profiled with the commonly used Infinium HumanMethylation450 BeadChip and the new enhancer-enriched Infinium MethylationEPIC Beadchip. Strong genetic effects (heritability > 0.4) were detected for 10% of sites and common genetic variants were identified to affect methylation levels at 22% of sites, from the 771,169 interrogated CpG sites across the genome. Second, I explored influences on the early-life methylome by comparing genome-wide DNA methylation profiles in naturally-conceived twins and twins conceived by in vitro fertilisation. Analysis of epigenetic profiles obtained by methylated DNA immunoprecipitation coupled with deep sequencing detected small changes at a gene previously associated with infertility, TNP1. Finally, I explored the effect of intrinsic factors on adult DNA methylation profiles, performing epigenome-wide studies of menopause and related phenotypes such as the use of hormone replacement therapy, which have metabolic consequences in middle-aged women. Epigenetic changes at seven CpG sites were associated with hormone replacement therapy. In summary, the results presented in this thesis give insights into genetic and specific environmental influences on the human epigenome.

The major focus of this thesis will concern the consequences and downstream impacts of obesity and adiposity-related traits on the human body by utilising RNA-sequencing measurements from three primary tissues (Subcutaneous adipose tissue, Whole blood & Skin) and one cell line, LCLs (Lymphoblastoid cell lines). I will discuss how we can use gene expression and population genetic variation to understand the heterogeneous nature of obesity outcome in the population and to uncover the complex relationship between the eects of obesity on peripheral tissue biology, the environment and the consequences of obesity on gene regulation. First, I will examine the extent of gene expression association measured in peripheral tissues to multiple cardio-metabolic, hormonal and adiposity related measurements. I will characterise the heritability of gene expression in these four sources and discuss the tissue specicity of both trait associations and genetic eects on gene expression. Second, I will describe how BMI can act as a potent modier of gene expression in adipose tissue by modelling BMI as an exposure/environment to detect and for the rst time replicate BMIdependent eQTLs (G BMI) that are specic to adipose tissue. Lastly, I will explore the cell type heterogeneity of adipose tissue, a pertinent problem for many investigators performing gene expression based studies in bulk complex tissues. I will show how many BMI gene expression associations are driven by macrophage heterogeniety amongst samples, that cell type variability is heritable and describe examples of cis-eQTLs driven by macrophage proportion in adipose tissue.

Metabolomics is an exciting area of research obtaining the chemical and metabolic signals from biosamples. Food consumption is strongly linked to metabolism making the metabolome an ideal phenotype for identifying biomarkers and helping refine and explore diet-disease associations. However, the metabolome is highly complex and influenced by many factors, such as age, disease, genetics and gut microbiota. Discordant monozygotic twins may provide a strong model for confirming association findings as they are matched for age, sex and the baseline genetic sequence. In this thesis, I explored the potential and applicability of the metabolome in nutritional research in two main areas: for identifying biomarkers of food exposure and further investigating the relationship of food intake with indicators of health. Firstly, I examined metabolomics profiles associated with self-reported food intakes and dietary patterns and confirmed these associations using the co-twin control method. I then identified top metabolite markers of food group intakes, and created and tested metabolite scores using multiple-methods. I searched for markers of gut microbiome diversity, an emerging indicator of health, by examining the association of alpha-diversity with blood metabolomics profiles and the relationship with diet and the metabolic syndrome. In the final chapter, I created a dietary score predictive of visceral fat mass, a strong risk factor for cardio-metabolic disease, and examined the degree to which the relationship between diet and visceral fat mass is mediated by associated metabolites and microbiome taxa. Throughout each chapter I used discordant monozygotic twins to validate top results. Overall, my findings show that metabolomics is a highly versatile tool for advancing nutrition research from biomarker identification through to untangling the impact of dietary exposures on indicators of metabolic health and the gut microbiome.

A complex interaction of environmental factors, stochastic events, and genetic susceptibility can lead to cancer development. The aim of this thesis is to investigate the DNA methylome for cancer, cancer risk, and prediction potential. Studies were performed in peripheral blood to explore systemic changes associated with cancer, and in skin for an in-depth view of total body naevus count, the strongest risk factor for melanoma. Peripheral blood DNA methylomes of 41 cancer-discordant female monozygotic (MZ) twin-pairs were assessed for changes associated with any cancers. The epigenome-wide association study (EWAS) identified one genome-wide significant and several suggestive differential methylated positions (DMPs), three of these showed predictive biomarker potential (near SASH1, COL11A2, and LINC00340). Early breast cancer specific DNA methylation changes were identified in peripheral blood obtained prior to diagnosis. The DNA methylomes were assessed by two genome-wide DNA methylation techniques in a total of 28 breast cancer discordant MZ twin-pairs. Three novel significant breast-cancer differential methylated regions (DMRs) were identified (in MECOM, PCGF3, and near ELN) that were suggestive of predictive biomarker potential. Skin DNA methylomes were investigated in association with the number of naevi across the body in 322 female individuals. Three genome-wide significant DMPs were identified in novel genes METRNL, C15orf48, and ARRDC1. Suggestive results included CTC1 and RAF, which are known genes involved in naevi predisposition and melanoma progression. Approximately half of the 48 suggestive DMRs were correlated with gene expression in cis. Overall, DNA methylation changes related to cancer, pan-cancer and breast cancer specifically, as well as with the melanoma risk factor naevus count were identified. These loci are excellent candidates for further research into their potential as biomarkers or risk factor biological mechanisms in cancer.

Ageing is a complex process and is the strongest risk factor for many diseases. To elucidate processes underlying biological ageing, many studies have investigated the associations of individual metabolites or genes with age and age-related phenotypes. While these studies improved our understanding of the pathogenic mechanisms for many diseases, they have their limitations. Analysing phenotypes separately in classical association studies typically neglects relationships and interactions between and across different phenotypes (e.g. comorbidities), genes (e.g. pleiotropy), and metabolites (e.g. through shared biochemical pathways). The aim of this thesis was to better understand age-related phenotypes and their interdependencies through exploring shared underlying processes. To this end, I first identified biomarkers of ageing and age-related diseases, focusing on chronic kidney disease, using metabolomics and glycomics technologies. I identified several metabolites associated with leukocyte telomere length, a common marker of biological ageing. Then I investigated the associations of molecular phenotypes with renal disease. Analysing metabolomic profiles associated with renal function in diabetic and non-diabetic cohorts illustrated similarities between the different aetiologies of kidney disease, such as the lack of renal conversion of amino acids, but also differences, particularly of lipid and energy metabolism. Subsequent analyses identified changes of Immun-oglobin G glycosylation as a novel inflammatory pathway involved in renal disease. Then, I assessed the potential of the faecal metabolome as a functional readout of the gut microbial community to investigate its association with biological ageing. While faecal metabolites were only moderately associated with age and renal function, they showed great potential as novel profiling method for studying the microbiome, particularly with respect to obesity. Next, I integrated metabolomics data from plasma, urine, and saliva to model cross-fluid metabolism individually for kidney disease patients and healthy controls. By comparing both models, I identified metabolic key processes impaired in kidney disease. Finally, I integrated metabolomic and glycomic biomarkers of ageing with other omics markers as well as extensive phenotypic data to investigate their multivariate interdependencies, underlying the comorbidities of age-related diseases. This comprehensive integration of age-related phenotypes highlighted several molecular mechanisms that potentially cause the joint occurrence of diseases with age. Con-sidering the complex aetiologies of different diseases and their dependencies will be needed to facilitate personalised healthcare. In conclusion, I have shown the future potential of sys-tems and network biology approaches for understanding disease mechanisms and precision medicine.

Epigenetics and metabolomics are rapidly growing areas of research, in part due to recent advances in technology that have allowed for a wide coverage of the human genome. Metabolites are small compounds present in cell and body fluids, and are involved in biochemical processes of the cell. Quantitative trait loci associated with levels of individual metabolites (mQTLs) have been identified from numerous metabolome GWAS. Here, I analysed metabolite levels in twins with the aim of identifying genetic variants that influence metabolomic traits (mQTLs) using two different metabolomics platforms, and consequently compared the results to report stable metabolites on both technologies to ultimately enable combining metabolite profiles across these two platforms. DNA methylation is a biochemical process that is vital for mammalian development. It is present throughout the genome and is the most extensively studied epigenetic mark. Previous studies have explored the heritability of DNA methylation and have identified methylation QTLs (meQTL). Here, I identified meQTLs with the goal of assesing the evidence of genetic effects influence not only DNA methylation levels, but also variability by using MZ-twin discordance as a measure of variance. Epigenetic mechanisms and metabolomic profiles have both been shown to play a role in gene expression and susceptibility for complex human disease. Here, I analysed the association between type 2 diabetes and metabolomic and epigenetic datasets and combined the data to identify connections between these levels of biological data at genetic variants linked to type 2 diabetes to gain more insight into the disease susceptibility and progression. Overall, the results confirmed previous findings of strong genetic influences on metabolites and extend current knowledge about genetic effects underlying several biochemical pathways. Additionally, the results also showed genetic influences on DNA methylation, and give insights into mechanisms by which genetic impacts epigenetic processes. Lastly, the findings show that specific genetic susceptibility variants for type 2 diabetes can also impact epigenetic and metabolomics profiles, and can help improve our understanding of the disease etiology.

The human microbiota is defined as the collective microbes living on and in the human body. Microbes can confer many benefits to the host including pathogen protection, complex carbohydrate breakdown and essential compound biosynthesis. Imbalances in the microbiome have been linked to disease, in particular obesity. The aim of this thesis is to explore the genetic and environmental effects on the human gut microbiome in twins, and characterise the microbiome species related to obesity. The primary dataset included gut microbiome 16s profiles of 982 twins from the TwinsUK cohort with extensive phenotype data. I first explored the heritability of the gut microbiome and obesity using twin modelling. The results revealed several heritable families, including the most heritable family, Christensenellaceae, which was also associated with lean BMI and measures of adiposity, including visceral fat. In addition, I found further associations between Christensenella and immune-related phenotypes such as white blood cell count, as well as host immune genetic variants. The next major analysis focused on identifying gut bacterial species that significantly differed between obese and lean individuals. I refer to these analyses as a Microbiome-Wide Association Study (MiWAS). The majority of the significant MiWAS results in obesity were obtained for members of the Ruminococcaceae and Lachnospiraceae. A candidate gene analysis was performed which aimed to identify obesity-associated human variants linked to these microbes. The strongest association was obtained between Lachnospiraceae and a variant within the human RPTOR gene, which controls the insulin-signalling pathway in response to nutrient availability by either binding or dissociating from MTOR. This finding and further associations of Lachnospiraceae with insulin, diet and cholesterol measures implicate a role for Lachnospiraceae in insulin resistance as well as obesity. The final project explored the association between the gut microbiome and metabolites in serum, plasma and faeces. First I identified associations between serum and plasma metabolites. The peak serum & plasma metabolite signals were obtained with the metabolites palmitate and cholesterol, which both play an important role in human metabolic health. The top associations were with the family Christensenellaceae indicating the importance of this microbial family not only in obesity, but also in extended metabolic phenotypes. Next, I obtained strong signals both between adiposity and faecal metabolites as well as with 16S microbial profiles. This indicates a useful role for faecal metabolomics as well as serum/plasma metabolomics in future phenotypic analyses. In conclusion, my thesis explored the genetic and environmental basis of the human microbiome in twins, identifying host genetic influences on the gut microbiome. The project also aimed to characterise the human gut microbiome in obesity. The results confirm and extend previous findings of obesity-related microbiome profiles, and identify interesting host genes that may provide a starting point to understand the interaction of host genetics with the gut microbiome in human metabolism and obesity.

Epigenome-wide association scans (EWAS) of human complex traits are a rapidly growing area of research, in part due to recent advances in technology that have allowed for a deeper coverage of the human methylome. One of the unique features of the human methylome is that it is dynamic and previous studies have shown that age can have a strong impact on DNA methylation patterns. The dynamic nature of DNA methylation also influences EWAS methodology, both from a statistical and biological perspective. In this thesis, I explored EWAS methods and applications to ageing and age-related phenotypes. Firstly, I estimated EWAS power under several simulation scenarios and study designs, and my results suggested that the majority of recent EWAS studies lack statistical power to detect small DNA methylation effect sizes. I then applied EWAS to identify differential methylation CpG sites associated with three phenotypes, including ageing, birth weight and smoking. One of the novel findings from this thesis was that hundreds of genome-wide significant ageing-related hypermethylated regions were identified across multiple tissues in twins. These findings confirm and extend previous work showing that ageing has a strong underlying effect on DNA methylation. Birth weight did not yield significant differential methylation sites, which may be partly explained by low power to detect modest methylation effects. Smoking is a well-known environmental risk factor for disease, and my analyses identified novel impacts of smoking on DNA methylation patterns in adipose tissue, which are of interest to cardiovascular and metabolic disease. I further explored the impacts of smoking by integrating DNA methylation and gene expression profiles in adipose tissue and in whole blood. In addition to identifying novel results, my findings also confirmed that the AHRR and F2RL3 genes showed stable and consistent changes related to smoking in both DNA methylation and gene expression profiles across tissues. My findings explored methodological issues in genome-wide methylation studies and showed that age and smoking have a strong and reproducible effect on DNA methylation across tissues in humans, which suggests that these factors should always be included as covariates in EWAS of human complex traits.

Age-related hearing impairment (ARHI) affects 46% of the population over the age of 48 and with increasing life expectancy in Western nations, this incidence is likely to rise. The causes for this disorder are still poorly understood but there is known to be a heritable component of around 65%. In addition, epigenetic regulation of gene expression changes over time and may explain many age-related traits. The basis of this research was to explore genetic and epigenetic factors in ARHI to understand better the mechanisms involved in its pathology. Hearing data were obtained from female volunteers of the TwinsUK register using the gold standard measure, air-conduction pure-tone audiometry and a web-based speechin-noise perception test. The prevalence of ARHI in TwinsUK was comparable to previous reports. Heritability estimates based on the classical twin model confirmed a moderate heritability of hearing ability in TwinsUK, supporting the use of this sample in genetic association studies. Genome-wide association with hearing ability was performed in TwinsUK but no genome-wide significant polymorphisms were identified possibly due to inadequate sample size. Accordingly, the data were combined with existing genome-wide association studies (GWAS) of hearing function from the G-EAR consortium. This meta-analysis resulted in a genome-wide significant association with an exonic SNP of the SIK3 gene encoding salt-inducible kinase 3, a novel gene reported to regulate metabolism and skeletal development via HDAC4. Immunohistochemistry of sik3 in mouse models confirmed striking expression profiles A small epigenome-wide association study of hearing ability in TwinsUK (n=115) revealed an epigenome-wide significant association with a probe in the promoter region of TCF25. Epigenome-wide associations at two highly associated probes (TCF25 and POLE) were replicated in an independent sample from TwinsUK (n=203). DNA methylation at these genes was negatively correlated with expression of the same, indicating gene expression repression by DNA methylation. Furthermore, using identical twins discordant for hearing loss, differentially methylated regions were found at genes ACP6 and CCNDBP1. This research supports a role of common genetic variants in ARHI, including the novel association with SIK3, which may be essential for healthy hair cell development and maintenance of spiral ganglion cells with increased age. Differentially methylated in hair cells and spiral ganglion cells of mouse cochlea, validating a putative function of SIK3 in hearing ability. regions were significantly associated with ARHI and showed an effect on gene expression despite small sample size, supporting a role of epigenetic modifications in ARHI. This research is the first to report genome-wide significant association with SIK3 and epigenome-wide significant associations at TCF25 with ARHI, which may shed light on the pathways involved in this disabling condition.